Marketing Agent — Baltimore Real Estate Lead Gen

Autonomous multi-agent marketing system for Nick Jungmarker | REALTOR® | Long & Foster Real Estate.

Generates daily Facebook/Instagram content, scores incoming leads, writes follow-up messages, and analyzes Meta ad performance — all tuned to the Baltimore rowhouse market and first-time buyer audience.

---

Architecture

orchestrator.py          ← Lead agent that runs the daily pipeline
│
├── agents/
│   ├── content_agent.py       ← Facebook + Instagram post generation
│   ├── ad_strategy_agent.py   ← Weekly Meta campaign targeting + budget strategy
│   ├── lead_scorer.py         ← Scores leads 0-10 from AgentOS
│   ├── followup_agent.py      ← Personalized SMS/email/voicemail scripts
│   ├── analytics_agent.py     ← Meta Ads API performance reports
│   └── market_agent.py        ← Baltimore market context for content
│
├── tools/
│   ├── meta_api.py            ← Facebook + Instagram Graph API client
│   ├── agentos_client.py      ← AgentOS (fsbo-agent) API wrapper
│   └── file_store.py          ← Draft file + log management
│
├── output/
│   ├── drafts/YYYY-MM-DD.md   ← Daily review file (human-readable)
│   └── logs/YYYY-MM-DD.log    ← Machine log
│
├── config.py                  ← Brand constants, API keys, settings
└── main.py                    ← CLI entry point

---

Setup

# Clone the repo
git clone https://github.com/jungmarker/marketing-agent.git
cd marketing-agent

# Install dependencies
pip install -r requirements.txt

# Configure environment
cp .env.example .env
# Edit .env with your API keys

Required env vars

Key	Where to get it
ANTHROPIC_API_KEY	console.anthropic.com
FACEBOOK_ACCESS_TOKEN	Meta Business Suite → Token Tools
FACEBOOK_PAGE_ID	Your Facebook page URL or Settings
INSTAGRAM_ACCOUNT_ID	Meta Business Suite → Instagram account

---

Usage

# Run the full daily pipeline
python main.py

# Run as a specific day (to generate Monday's strategy on any day)
python main.py --day Monday

# Individual agents
python main.py --content       # Generate today's social content only
python main.py --score         # Score leads from AgentOS only
python main.py --strategy      # Generate weekly ad strategy only
python main.py --analytics     # Force weekly analytics report

# Start daily scheduler (runs at 9am)
python main.py --schedule

---

Posting Modes

Set in .env:

Mode	Behavior
POSTING_MODE=draft	Saves to output/drafts/YYYY-MM-DD.md for review before posting
POSTING_MODE=auto	Posts directly to Facebook + Instagram (requires valid tokens)

Start in draft mode. Review output/drafts/ before switching to auto.

---

Sub-Agent Overview

content_agent — Daily Social Content

• Reads the content calendar (Monday=educational, Tuesday=carousel, etc.)
• Picks a Baltimore neighborhood to spotlight
• Generates Facebook post + Instagram caption + visual description
• Bilingual — writes in Spanish for Spanish-speaking audiences
• Always appends compliance footer

ad_strategy_agent — Weekly Meta Strategy (Mondays)

• Recommends budget split between Campaign A (FTB guide) and Campaign B (seller valuation)
• Generates 3 headline variations + 3 primary text variations
• Proposes A/B test ideas
• Stays within Housing Special Ad Category constraints (no demographic targeting)

lead_scorer — Lead Prioritization

• Pulls new leads from AgentOS API
• Scores 0–10 on: contact quality, timeline, motivation, pre-qualification, source
• Returns tiered list: hot / warm / medium / cold

followup_agent — Personalized Outreach

• Generates SMS (<155 chars), email subject + body, voicemail script
• Uses lead's name, source, and message to personalize
• Writes in Spanish if lead appears Spanish-speaking
• Stages: initial → day3 → day7 → day14 → day30 → reactivation

analytics_agent — Meta Ad Performance (Saturdays)

• Fetches campaign spend, impressions, CTR, leads, CPL from Meta Ads API
• Returns plain-English summary + wins/concerns/recommendations
• Uses mock data until real FACEBOOK_ACCESS_TOKEN is configured

market_agent — Baltimore Market Context

• Generates weekly Baltimore market narrative (prices, inventory, seasonality)
• Spotlights 2 neighborhoods
• Produces content hooks for social media
• Feeds context into content + strategy agents

---

Content Calendar

Day	Theme	Format
Monday	Educational	Video script
Tuesday	Carousel	Neighborhood spotlight
Wednesday	Seller/Market	Static image
Thursday	Neighborhood	Reel
Friday	Personal	Static image
Saturday	Market data	Carousel
Sunday	Community	Static image

---

Compliance

Every post auto-includes:

Nick Jungmarker | REALTOR® | Long & Foster Real Estate
(301) 875-7182 | jungmarker.com
Equal Housing Opportunity 🏠

All Meta ads use Special Ad Category: Housing — no demographic targeting or exclusions.

---

Integration with AgentOS

When the fsbo-agent backend is running locally at http://localhost:8000:

• lead_scorer pulls and scores real leads
• followup_agent pulls lead details for personalization
• agentos_client can write follow-up notes back to lead records

---

Remote Access

To continue work remotely:

1. GitHub — this repo. Clone on any device.
2. GitHub Codespaces — open the repo → Code → Codespaces → New. Full Claude Code access from browser/phone.
3. SSH to Mac — install Tailscale on your Mac, then SSH in from any device.

---

Nick Jungmarker | REALTOR® | Long & Foster Real Estate | Maryland Licensed Equal Housing Opportunity